两列上的 PIVOT 运算符

Question

我对 SQL 很陌生，更不用说在 SSMS 中做 PIVOTS。在做了一些研究之后，我很难为两列做枢轴。

我查询时的数据本身是：

SELECT
   INSTRUMENT_ID,
   VALUE_TYPE_CODE,
   ORIGINAL_TRANSACTION_VALUE,
   VALUE_IN_HOME_CURRENCY

FROM
   FACT_TABLE

INSTRUMENT_ID | VALUE_TYPE_CODE | ORIGINAL_TRANSACTION_VALUE | VALUE_IN_HOME_CURRENCY
ABC           |TOT_AMT          |   3                        |  5.7702
ABC           |AMT_AVAL         |   1                        |  1.9234
ABC           |DRWN_AMT         |   2                        |  3.8468
ABC           |MTD_DRWN         |   2                        |  3.8468

以ORIGINAL_TRANSACTION_VALUE

为中心

SELECT *
FROM
(
  SELECT INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       ORIGINAL_TRANSACTION_VALUE,
       VALUE_IN_HOME_CURRENCY
  FROM FACT_TABLE
)
AS Table
PIVOT (MAX (ORIGINAL_TRANSACTION_VALUE) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                                [AMT_AVAL], 
                                                                [DRWN_AMT], 
                                                                [MTD_DRWN])) as piv

结果如预期：

INSTRUMENT_ID | TOT_AMT | AMT_AVAL | DRWN_AMT | MTD_DRWN
ABC           |3        | 1        | 2        | 2

但是，我正在尝试生成以下内容 - 请注意每个 VALUE_TYPE_CODE 类别的交替 ORIGINAL_TRANSACTION_VALUE 和 VALUE_IN_HOME_CURRENCY 值

INSTRUMENT_ID |TOT_AMT  |TOT_AMT |AMT_AVAL |AMT_AVAL |DRWN_AMT |DRWN_AMT |MTD_DRWN |MTD_DRWN
ABC           | 3       |5.7702  |  1      |1.9234   |2        |3.8468   |  2      |3.8468

所以我尝试了从这里找到的以下内容。它看起来很简单，而且我还是新 SQL，所以对我来说很有意义。复制类别值，否则 PIVOT 将无法正常工作。

SELECT *
FROM
(
  SELECT INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       VALUE_TYPE_CODE+'1' as VALUE_TYPE_CODE1,
       ORIGINAL_TRANSACTION_VALUE,
       VALUE_IN_HOME_CURRENCY
  FROM FACT_TABLE
)
AS Table
-- Pivot on ORIGINAL_TRANSACTION_VALUE
PIVOT (MAX (ORIGINAL_TRANSACTION_VALUE) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                                 [AMT_AVAL], 
                                                                 [DRWN_AMT], 
                                                                 [MTD_DRWN])) as piv
-- Pivot on VALUE_IN_HOME_CURRENCY
PIVOT (MAX (VALUE_IN_HOME_CURRENCY) FOR  VALUE_TYPE_CODE1 IN ([TOT_AMT1], 
                                                              [AMT_AVAL1], 
                                                              [DRWN_AMT1], 
                                                              [MTD_DRWN1])) as piv1

不幸的是，这并没有按照我希望的方式工作，结果如下：

INSTRUMENT_ID |TOT_AMT  |TOT_AMT |AMT_AVAL |AMT_AVAL |DRWN_AMT |DRWN_AMT |MTD_DRWN |MTD_DRWN
ABC           | 3       |5.7702  |  NULL   |NULL     |NULL     |NULL     |NULL     |NULL   
ABC           | NULL    |NULL    |  1      | 1.9234  |NULL     |NULL     |NULL     |NULL 
ABC           | NULL    |NULL    |  NULL   |NULL     |2        |3.8468   |NULL     |NULL  
ABC           | NULL    |NULL    |  NULL   |NULL     |NULL     |NULL     |  2      |3.8468

我想知道是否有人能指出我做错了什么？有没有稍微简单一点的方法？本身可能不是一种更简单的方法，而是一种通过一些解释分解的解决方案。我在这里查看了一些解决方案，其中很多只是让我头疼，因为我还是新手。

我非常感谢任何帮助。谢谢。

Answer 1

主要问题是您为数据透视语句提供的列比您要透视的列多。 piv 不需要 VALUE_IN_HOME_CURRENCY 用于其枢轴操作，piv1 不需要 ORIGINAL_TRANSACTION_VALUE。

这个问题有几种解决方案：

1. 将group by 和一些聚合添加到您当前的解决方案中。
2. 将枢轴操作拆分为两个单独的操作，然后在 INSTRUMENT_ID 上重新加入。
3. 取消透视原始数据，使ORIGINAL_TRANSACTION_VALUE 和VALUE_IN_HOME_CURRENCY 成为一列，然后一次性旋转为8 列。在单个列中获取所有值确实需要所有值的兼容数据类型。

样本数据

create table FACT_TABLE
(
   INSTRUMENT_ID nvarchar(3),
   VALUE_TYPE_CODE nvarchar(10),
   ORIGINAL_TRANSACTION_VALUE int,
   VALUE_IN_HOME_CURRENCY money,
);

insert into FACT_TABLE (INSTRUMENT_ID, VALUE_TYPE_CODE, ORIGINAL_TRANSACTION_VALUE, VALUE_IN_HOME_CURRENCY) values
('ABC', 'TOT_AMT' , 3, 5.7702),
('ABC', 'AMT_AVAL', 1, 1.9234),
('ABC', 'DRWN_AMT', 2, 3.8468),
('ABC', 'MTD_DRWN', 2, 3.8468);

解决方案 1

SELECT piv1.INSTRUMENT_ID,
       max(piv1.TOT_AMT)   as TOT_AMT,
       max(piv1.TOT_AMT1)  as TOT_AMT1,
       max(piv1.AMT_AVAL)  as AMT_AVAL,
       max(piv1.AMT_AVAL1) as AMT_AVAL1,
       max(piv1.DRWN_AMT)  as DRWN_AMT,
       max(piv1.DRWN_AMT1) as DRWN_AMT1,
       max(piv1.MTD_DRWN)  as MTD_DRWN,
       max(piv1.MTD_DRWN1) as MTD_DRWN1
FROM
(
  SELECT INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       VALUE_TYPE_CODE+'1' as VALUE_TYPE_CODE1,
       ORIGINAL_TRANSACTION_VALUE,
       VALUE_IN_HOME_CURRENCY
  FROM FACT_TABLE
)
AS T
-- Pivot on ORIGINAL_TRANSACTION_VALUE
PIVOT (MAX (ORIGINAL_TRANSACTION_VALUE) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                                 [AMT_AVAL], 
                                                                 [DRWN_AMT], 
                                                                 [MTD_DRWN])) as piv
-- Pivot on VALUE_IN_HOME_CURRENCY
PIVOT (MAX (VALUE_IN_HOME_CURRENCY) FOR  VALUE_TYPE_CODE1 IN ([TOT_AMT1], 
                                                              [AMT_AVAL1], 
                                                              [DRWN_AMT1], 
                                                              [MTD_DRWN1])) as piv1
group by piv1.INSTRUMENT_ID

解决方案 2

with cte_piv as
(
  -- Pivot on ORIGINAL_TRANSACTION_VALUE
  select piv.INSTRUMENT_ID,
         piv.TOT_AMT,
         piv.AMT_AVAL,
         piv.DRWN_AMT,
         piv.MTD_DRWN
  from ( SELECT INSTRUMENT_ID, VALUE_TYPE_CODE, ORIGINAL_TRANSACTION_VALUE
         FROM FACT_TABLE ) T
  PIVOT (MAX (ORIGINAL_TRANSACTION_VALUE) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                                 [AMT_AVAL], 
                                                                 [DRWN_AMT], 
                                                                 [MTD_DRWN])) as piv
),
cte_piv1 as
(
  -- Pivot on VALUE_IN_HOME_CURRENCY
  select piv1.INSTRUMENT_ID,
         piv1.TOT_AMT,
         piv1.AMT_AVAL,
         piv1.DRWN_AMT,
         piv1.MTD_DRWN
  from ( SELECT INSTRUMENT_ID, VALUE_TYPE_CODE, VALUE_IN_HOME_CURRENCY
         FROM FACT_TABLE ) T
  PIVOT (MAX (VALUE_IN_HOME_CURRENCY) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                              [AMT_AVAL], 
                                                              [DRWN_AMT], 
                                                              [MTD_DRWN])) as piv1
)
select p.INSTRUMENT_ID,
       p.TOT_AMT,
       p1.TOT_AMT,
       p.AMT_AVAL,
       p1.AMT_AVAL,
       p.DRWN_AMT,
       p1.DRWN_AMT,
       p.MTD_DRWN,
       p1.MTD_DRWN
from cte_piv p
join cte_piv1 p1 on p1.INSTRUMENT_ID = p.INSTRUMENT_ID

解决方案 3

with cte as
(
  SELECT INSTRUMENT_ID,
         VALUE_TYPE_CODE,
          -- all values require a compatible type for a pivot
         convert(money, ORIGINAL_TRANSACTION_VALUE) as ORIGINAL_TRANSACTION_VALUE,
         VALUE_IN_HOME_CURRENCY
 FROM FACT_TABLE
),
cte2 as
(
  select up.INSTRUMENT_ID,
         up.VALUE_TYPE_CODE + '_' + left(up.ValueType, 4) as NEW_CODE,
         up.Value
  from cte
  unpivot (Value for ValueType in (ORIGINAL_TRANSACTION_VALUE, VALUE_IN_HOME_CURRENCY)) up
)
select p.INSTRUMENT_ID,
       convert(int, p.TOT_AMT_ORIG)  as TOT_AMT,
       p.TOT_AMT_VALU                as TOT_AMT1,
       convert(int, p.AMT_AVAL_ORIG) as AMT_AVAL,
       p.AMT_AVAL_VALU               as AMT_AVAL1,
       convert(int, p.DRWN_AMT_ORIG) as DRWN_AMT,
       p.DRWN_AMT_ORIG               as DRWN_AMT1,
       convert(int, p.MTD_DRWN_ORIG) as MTD_DRWN,
       p.DRWN_AMT_VALU               as MTD_DRWN1,
       convert(int, p.MTD_DRWN_ORIG) as MTD_DRWN,
       p.MTD_DRWN_VALU               as MTD_DRWN1
from cte2
pivot (max(Value) for NEW_CODE in ([TOT_AMT_ORIG],  [TOT_AMT_VALU],
                                   [AMT_AVAL_ORIG], [AMT_AVAL_VALU],
                                   [DRWN_AMT_ORIG], [DRWN_AMT_VALU],
                                   [MTD_DRWN_ORIG], [MTD_DRWN_VALU])) p

Fiddle 1 = 解决方案 1 和 2
Fiddle 2 = 解决方案 3

Answer 2

此答案适用于希望使其动态化的观众。 [列值是动态的]

对于 MSSQL SERVER，

DECLARE @cols1 AS NVARCHAR(MAX),@cols2 AS NVARCHAR(MAX),@selectcolumn1 AS NVARCHAR(MAX),@selectcolumn2 AS NVARCHAR(MAX),
   @selectcolumn AS NVARCHAR(MAX), @query  AS NVARCHAR(MAX);
    SET @cols1 = STUFF((SELECT distinct ',' + QUOTENAME(f.VALUE_TYPE_CODE) 
            FROM FACT_TABLE f
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'') 
    SET @cols2 = STUFF((SELECT distinct ',' + QUOTENAME(f.VALUE_TYPE_CODE+'1')
            FROM FACT_TABLE f
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'') 

    SET @selectcolumn1 = STUFF((SELECT distinct ',MAX(piv1.' + f.VALUE_TYPE_CODE+') as '+ f.VALUE_TYPE_CODE+'_OTV'
        FROM FACT_TABLE f
        FOR XML PATH(''), TYPE
        ).value('.', 'NVARCHAR(MAX)') 
    ,1,1,'') 

    SET @selectcolumn2 = STUFF((SELECT distinct ',MAX(piv1.' + f.VALUE_TYPE_CODE+'1) as '+ f.VALUE_TYPE_CODE+'_HC'
        FROM FACT_TABLE f
        FOR XML PATH(''), TYPE
        ).value('.', 'NVARCHAR(MAX)') 
    ,1,1,'')

  set @selectcolumn=concat(@selectcolumn1,',',@selectcolumn2)

set @query = 'SELECT piv1.INSTRUMENT_ID,
       '+@selectcolumn+'
FROM
(
  SELECT INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       VALUE_TYPE_CODE+''1'' as VALUE_TYPE_CODE1,
       ORIGINAL_TRANSACTION_VALUE,
       VALUE_IN_HOME_CURRENCY
  FROM FACT_TABLE
)
AS T
PIVOT (MAX (ORIGINAL_TRANSACTION_VALUE) FOR  VALUE_TYPE_CODE IN ('+@cols1+')) as piv
PIVOT (MAX (VALUE_IN_HOME_CURRENCY) FOR  VALUE_TYPE_CODE1 IN ('+@cols2+')) as piv1
                                                              
group by piv1.INSTRUMENT_ID'

execute(@query)

对于 MYSQL 服务器

SET @col1 = NULL;
SET @col2 = NULL;
SET @sql = NULL;
SELECT
  GROUP_CONCAT(DISTINCT
    CONCAT(      
 
       'Max(IF(f.VALUE_TYPE_CODE = ''',VALUE_TYPE_CODE,''', ORIGINAL_TRANSACTION_VALUE, Null)) as ',VALUE_TYPE_CODE,'_OT'
    )
  ) INTO @col1
FROM FACT_TABLE;

SELECT
  GROUP_CONCAT(DISTINCT
    CONCAT(      
 
       'Max(IF(f.VALUE_TYPE_CODE1 = ''',VALUE_TYPE_CODE,'1'', VALUE_IN_HOME_CURRENCY, Null)) as ',VALUE_TYPE_CODE,'_HC'
    )
  ) INTO @col2
FROM FACT_TABLE;



SET @sql = CONCAT('SELECT 
    f.INSTRUMENT_ID,  ', CONCAT(@col1,',',@col2) , 
    ' FROM  (SELECT INSTRUMENT_ID,VALUE_TYPE_CODE,CONCAT(VALUE_TYPE_CODE,''1'') AS VALUE_TYPE_CODE1,ORIGINAL_TRANSACTION_VALUE,VALUE_IN_HOME_CURRENCY FROM FACT_TABLE) f
   Group by f.INSTRUMENT_ID');
  
PREPARE stmt FROM @sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;
   

Answer 3

PIVOT 要记住的是，它会在该点转换整个结果集。 PIVOT 中未提及的所有列都有效地用于将结果分组到行中。

因此，在您的尝试中，您的第一个 PIVOT 根据 INSTRUMENT_ID 和 VALUE_IN_HOME_CURRENCY 对行进行分组，这就是它产生四行的原因。

解决方案是将原始结果集旋转两次，排除不必要的列，然后加入结果。

所以让我们使用公用表表达式来分解它：

WITH SourceData as (
    SELECT INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       ORIGINAL_TRANSACTION_VALUE,
       VALUE_IN_HOME_CURRENCY
    FROM FACT_TABLE
), OriginalPivotSource as (
    SELECT
       INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       ORIGINAL_TRANSACTION_VALUE
    FROM SourceData
), OriginalPivot as (
    SELECT
        *
    FROM OriginalPivotSource
    PIVOT (MAX (ORIGINAL_TRANSACTION_VALUE) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                                 [AMT_AVAL], 
                                                                 [DRWN_AMT], 
                                                                 [MTD_DRWN])) t
), HomePivotSource as (
    SELECT
       INSTRUMENT_ID,
       VALUE_TYPE_CODE,
       VALUE_IN_HOME_CURRENCY
    FROM
        SourceData
), HomePivot as (
    SELECT
        *
    FROM HomePivotSource
    PIVOT (MAX (VALUE_IN_HOME_CURRENCY) FOR  VALUE_TYPE_CODE IN ([TOT_AMT], 
                                                              [AMT_AVAL], 
                                                              [DRWN_AMT], 
                                                              [MTD_DRWN])) t
)
SELECT
    *
FROM
    OriginalPivot op
        FULL OUTER JOIN
    HomePivot hp
        on
           op.INSTRUMENT_ID = hp.INSTRUMENT_ID